How reliable is CPU_SEC for capacity planning? Is it reliable between different servers? There are cases where I may have dev,test environment and get asked for an estimate of CPUs needed for prod. If I run tests with what we think is a percentage of peak load in a test environment (since test will usually have less horsepower than a prod environment) can I spreadsheet that and project it to a prod environment ? Do you have an estimate for how much cpu power I need to reserve in the spreadsheet just for background processes? There are also cases where people want to consolidate many DBs to 1 Server and estimating this can be tricky. I tend to do alot of guessing. wondering if anyone has any half way decent spreadsheet formulas? On 1/28/14, Jonathan Lewis <jonathan@xxxxxxxxxxxxxxxxxx> wrote: > > You're thinking of my rough guide of : 10,000 buffer gets per second per > 100MHz of CPU. (I think 100MHz was the unit, but I can't find the note at > present). > I wasn't using that in this case (after all, we now know that the SQL was > doing a lot of sorting and it;'s also applying some analytics). > > This was simple: > EXECUTIONS = 789388 > CPU_SEC = 1719853.25 > > CPU per Exec = 2.1787 seconds per exec > > > Regards > Jonathan Lewis > http://jonathanlewis.wordpress.com > @jloracle > -- //www.freelists.org/webpage/oracle-l